Diversity in Fermi/GBM Gamma-Ray Bursts: New Insights from Machine Learning

Abstract Classification of gamma-ray bursts (GRBs) has been a long-standing puzzle in high-energy astrophysics. Recent observations challenge the traditional short versus long viewpoint, where long GRBs are thought to originate from the collapse of massive stars and short GRBs from compact binary mergers. Machine learning (ML) algorithms have been instrumental in addressing this problem, revealing five distinct GRB groups within the Swift Burst Alert Telescope (BAT) light-curve data, two of which are associated with kilonovae (KNe). In this work, we extend our analysis to the Fermi Gamma-ray Burst Monitor catalog and identify five clusters using unsupervised ML techniques, consistent with the Swift/BAT results. These five clusters are well separated in the fluence-duration plane, hinting at a potential link between fluence, duration, and complexities (or structures) in the light curves of GRBs. Further, we confirm two distinct classes of KN-associated GRBs. The presence of GRB 170817A in one of the two KN-associated clusters lends evidence to the hypothesis that this class of GRBs could potentially be produced by binary neutron star mergers. The second KN-associated GRB cluster could potentially originate from neutron star–black hole mergers. Future multimessenger observations of compact binaries in gravitational waves and electromagnetic waves can be paramount in understanding these clusters better.